The black-haired woman stares enigmatically at an unseen object to her right, bearing a striking resemblance to the person depicted in Pierre-Auguste Renoir’s painting Gabrielwhich Sotheby’s recently valued between £100,000-150,000.
However, art connoisseurs disagree on whether the work, which belongs to a private Swiss collector, is genuine. Now artificial intelligence has stepped in to help resolve the dispute, and the computer has assumed it’s likely a real Renoir.
AI is increasingly being used to assess whether valuable works of art are real or fake. Earlier this month, Art Recognition, the Swiss company that developed the technology, announced that it had concluded that the only Titian in Switzerland – a work titled Evening Landscape with Couple, owned by Kunsthaus Zurich – was probably not painted. by the 16th century Venetian artist.
Still, art experts have warned that the AI is only as good as the paintings it’s trained on. If they are fake or contain parts that have been updated, it can add even more uncertainty.
Art Recognition was approached about Renoir, titled Portrait de femme (Gabrielle), named after The Wildenstein Plattner Institute – one of two institutes to publish a comprehensive list of all of Renoir’s known works of art, known as a catalogs raisonnés – refused to include it in his list.
The company used photographic reproductions of 206 authentic paintings by the French Impressionist to teach its algorithm about his style, which to human observers is characterized by broken brushstrokes and bold combinations of complementary colors. To increase precision, it also split the images into smaller chunks and showed them to the algorithm, training it on a selection of paintings by artists with a similar style who were active around the same time as Renoir.
Based on this assessment, it was concluded that there was an 80.58% chance that Portrait de femme (Gabrielle) was painted by Renoir.
Carina Popovici, CEO of Art Recognition, believes this ability to quantify the level of uncertainty is important. She said Monday at a meeting on the use of forensics and technology in the art trade at the Art Loss Register in London: “Art owners are often told by connoisseurs that it is their ‘impression’ or ‘intuition’ that a painting is real or no, which can be very frustrating. They really appreciate that we are more accurate.”
Encouraged by this result, the owner of the painting approached another Parisian group of experts, GP.F. Dauberville & Archives Bernheim-Jeune, which are its own catalog raisonnés of works by Renoir. After requesting a scientific analysis of the pigments in the painting, they too concluded that it was a genuine Renoir.
Dr. Bendor Grosvenor, an art historian and presenter of BBC Four’s Britain’s Lost Masterpieces, was concerned that such technologies would devalue the contribution of experts in assessing a work of art’s authenticity.
“Until now, the methods used to ‘train’ the AI programs and the fact that they say they can judge an attribution based only on an iPhone photo are not impressive,” he said.
“The technology is particularly weak in its inability to take into account the condition of a painting – so many Old Master paintings are damaged and disfigured by layers of dirt and overpaint, making it difficult without forensic inspection to distinguish what is and what is not original is.
“If an appraiser of human art offered to provide a ‘certificate of authenticity’ costing thousands of dollars, based on nothing more than an iPhone photo and a partial knowledge of an artist’s body of work, they would be laughed at. “
Popovici agreed that the quality of the training dataset was essential, saying they made every effort to ensure they only use photos of authentic artwork. So far, they’ve trained their AI to recognize about 300 artists, including most of the French Impressionists and Old Masters.
“We understand that those in the know might feel threatened by this technology, but we’re not trying to get rid of them,” Popovici said.
“We really want to give them the opportunity to use this system to help them make a decision, maybe in cases where they’re not so sure. But for that to happen, they have to be open to this technology.”
Julian Radcliffe, the chair of the Art Loss Register, which maintains the world’s largest private database of stolen art, antiques and collectibles, said: “Artificial intelligence is playing an increasing role in helping authenticate art, but it needs to be linked to the expertise of connoisseurs who specialize in the artist, established science such as pigment analysis and provenance research.
“The advantage lies in the ability to provide yes/no answers to, for example, pattern analysis or matching, and to continuously improve, but the work must be interpreted by a human being who must have asked the right question.
“The quest for absolute certainty in authentication has not been and may never be reached – but we are getting closer.”